Guys, Made change to use new libvirt event impl, but as Dan said, we need more regression tests for that. I'm not particularly sure what are necessary.Searched buglist found nothing relevant. Do we have a regression list for this? I can examine them one by one.
2012/5/31 Saggi Mizrahi smizrahi@redhat.com
+1 to the not understanding +1 to the bug fixing
----- Original Message -----
From: "Dan Kenigsberg" danken@redhat.com To: "Royce Lv" lvroyce0210@gmail.com Cc: vdsm-devel@lists.fedorahosted.org Sent: Wednesday, May 30, 2012 10:43:54 AM Subject: Re: [vdsm] setgroup/setuid affects libvirt event loop results
in connection broken
On Wed, May 30, 2012 at 10:28:29PM +0800, Royce Lv wrote:
Guys, In debugging vdsm/supervdsm process I noticed that event loop in libvirtev.py will be affected by setgroups/setgid/setuid. Results in libvirtd close connection because the keepalive scheme fals.Follow up vdsm to libvirt calls fails because of broken pipe.(detail: https://bugzilla.redhat.com/show_bug.cgi?id=825634) I'm afraid this will make vdsm/libvirt connection vulnerable in some of the current use case,too. So I have plan: 1.we use default impl as https://bugzilla.redhat.com/show_bug.cgi?id=681885 mentioned 2.use the setKeepAlive api in libvirt (keepalive problem seems disappear with this) 3.start libvirt event loop in first time we get libvirt connection, not when the file first loaded . I'd appreciate ideas about this on how to address this.Thanks a lot!
I don't completely grasp points 2 and 3, but Bug 681885 - RFE: Replace python event loop code, with use of new libvirt event APIs is long due, please implement it!
Dan. _______________________________________________ vdsm-devel mailing list vdsm-devel@lists.fedorahosted.org https://fedorahosted.org/mailman/listinfo/vdsm-devel